3 evidence-linked departures
Anthropic, founded as a safety-focused lab, has two evidence-linked departures in this tracker. In February 2026, Head of Safeguards Research Mrinank Sharma resigned warning that 'the world is in peril' and described tension between Anthropic's safety goals and competitive pressure. In September, pre-training researcher Jacob Coxon resigned after concluding that neither Anthropic nor OpenAI was acting responsibly in the race toward self-improving AI. Coxon said Anthropic's safety work was sincere, but argued that competition would eventually force unacceptable trade-offs.